Outbound sequences that survive a real inbox
An outbound sales sequence is a set of decisions about attention: how many times you are willing to interrupt one person, over how long, on which channels, and on what condition you stop. The copy decides whether a touch lands. The architecture decides whether it ever happens, and whether you are still welcome by the time it does.
By Rahul Bageria, co-founder · Updated August 2026 · 21 min read
Written by operators who run email and LinkedIn sequences for seed-stage teams, not by a vendor selling a sequencer.
What an outbound sequence actually is
The tools cannot agree what to call this. Outreach, HubSpot and Apollo say sequence, Salesloft says cadence, lemlist and Instantly say campaign. It is one object with three names, decided by whoever named the database table.
An outbound sequence is the planned series of touches to one person over time: which channels, in what order, with what spacing, and on what condition it stops.
The distinction worth teaching is not sequence against cadence. It is the nesting, which the tools enforce whether you name it or not:
- 1A campaign is the container: one segment, one offer, one list.
- 2A sequence is what happens to one person inside it.
- 3A step is one touch, on one channel, with one job.
Run one big sequence across four segments and you are running one campaign with four wrong offers in it. You cannot read the result, because a segment that works and one that does not average out to mediocre.
Architecture versus content: the split that decides what you fix
Reply rate flat? Almost every founder rewrites the emails. Half the time the emails were fine and the sequence was sending them into a hole.
| Dimension | Architecture decides | Content decides |
|---|---|---|
| The question it answers | How often, over how long, on which channels, and when to stop. | What each touch actually says. |
| The failure it causes | Touches that never happen, or arrive when you are no longer welcome. | Touches that happen and get ignored. |
| The symptom on your screen | Interest that starts and dies. Opt-outs clustered on one step. | Flat replies from step one, across every angle you try. |
| The fix | Spacing, step jobs, channel mix, branching, exit rules. | The angle, the opening line, the ask. |
| What teams actually do | Inherit it from a template and never look at it again. | Rewrite it, again, on a Friday afternoon. |
The copy decides whether a touch lands. The architecture decides whether the touch ever happens, and whether you are still welcome by the time it does.
What a step is actually for
A step with no job is the step that says "just following up", and that step is why people mute you. There are about five jobs a step can do.
Open a reason to be in their inbox
Step one answers why now, not why you. Something observable happened, or a specific problem is live in their world this quarter. Every later step inherits the credibility of this one, which is why a weak opener cannot be repaired by a strong follow-up.
Add something step one did not say
A number, a comparable situation, a short proof, a different reading of their problem. If a step cannot carry information the previous one did not, cut it. The same message five times in escalating politeness is the most common broken sequence there is.
Get seen by a different filter
A channel switch is not a lift trick, it is insurance. Your email may have been filed, filtered or simply missed, and none of those are a decision about you. Changing surface gives the same message one honest second look.
Change the ask, not the volume
Make a no to the meeting stop being a no to everything. Ask who owns this instead. Ask whether it is worth revisiting next quarter. A smaller ask keeps a door open that a fourth meeting request quietly closes.
End it on purpose
The last step says you are stopping and leaves the door on the latch. Written well, an ending out-replies more pitching at the same point in the run. Written as a guilt trip about being ignored, it is the reason somebody marks you as spam instead.
How long a sequence should be
Length is a function of two things: how big the ask is, and how warm the relationship already is. Both of these are correct answers.
Somebody who already bought from you, or a champion who has landed somewhere new. The relationship does the work and the sequence only stops you forgetting, which is exactly how the champion tracking play runs it.
An account you actually want, and a buyer who owes you nothing. Salesloft's own cadence playbooks publish a C-level track measured in months rather than weeks, on the stated logic that you have not earned the meeting early and should not ask for it.
The "8 to 12 touches" printed on every ranking page is an average of those two answers, and an average of two correct answers is a third wrong one. Decide the ask and the temperature first. The number falls out of that, and then you check it against your send capacity before anything else.
Where "eight touches" actually came from
The number is real and traceable, which makes it more dangerous than an invented one. Here is what it is, and what it is not.
It takes an average of 8 touchpoints to get an initial meeting or other conversion with a new prospect.
Sellers, reporting on themselves
RAIN Group's research page says the figure comes from 489 sellers who prospect outbound. That is people recalling their own averages, not a system counting touches. Checked August 2026.
Any conversion, on an unstated date
Read the sentence again: touches to an initial meeting or other conversion, so a demo, a trial and a reply that went somewhere all count. Neither the blog post carrying it nor the research page says when the sellers were asked.
A description, not a design target
It describes what a panel of sellers said happened to them. Building your architecture on it is like setting your salary from a survey of what other people earn: interesting, and no substitute for the arithmetic in front of you.
Quote it as trivia if you like, but do not design against it. The number worth having is your own: how many touches your last ten meetings actually took, and where in the run the replies came back.
Spacing: front-load, then stretch
Touches close together early establish that a person exists. The same tightness in week four establishes that a robot exists.
Space it evenly and let it trail off
Day 1, 4, 7, 10, 13, 16, 19, 22, 25, 28
- ✕Every gap says the same thing
- ✕Week four reads as automation
- ✕Nothing marks the ending
Cluster, break, widen, close
Day 1, 3, 6 · break · Day 13, 20 · Day 34 close
- ✓The first week says a person is here
- ✓The break is designed, not accidental
- ✓The last step was written first
Salesloft's cadence playbooks build silence in deliberately and say why: to stop prospect fatigue and to keep the ramp of touches from running away mid-sequence. A break you designed reads completely differently from a gap you drifted into, and the reader can tell which one they are in. Spacing is days between touches; which hour a touch lands in is a separate and much smaller question.
Channel mix, and the limit LinkedIn will not publish
A second channel is insurance, not a multiplier. It exists so one filter, one inbox and one bad week cannot end the conversation on their own.
A different filter, a different attention pattern, and independence from a single deliverability event. Not a lift number, whatever the claims in circulation say.
Every channel adds a queue to maintain and a set of rules to obey. Two channels run properly beat three run badly, every time, at two people.
LinkedIn's own help center confirms a commercial use limit exists, says it cannot display how many searches or views remain, and cannot lift the limit on request. Its prohibited software page bans bots and browser extensions that add or download contacts or send messages, and warns they may stop working without notice. Checked August 2026. So every LinkedIn step in your sequence has to survive being done by hand, because one Tuesday it will be.
The day-by-day version lives next door
This guide teaches the cadence principle. The worked calendar, with both channels interleaved and every gap accounted for, is a separate piece, and so is the craft of the LinkedIn steps themselves.
Want the architecture built around your segments instead of a template?
Book a Fit CheckBranching: what happens when they do something
A sequence with no branches is a broadcast on a timer. Three branches cover almost everything a two-person team needs.
-
1
They engage, but nobody replies
Tighten the next gap and change the ask. Do not add three more steps. Engagement without a reply usually means the offer is close but the request is too big.
-
2
Somebody else at the account engages
The thread moves to them, and the original contact drops to a lighter track rather than being dropped entirely. This is the branch that pays, and it is the one most sequences do not have.
-
3
Nothing happens at all
A non-reply is a routing decision, not a rejection. If the account still fits, this person, this angle or this month was wrong. End the sequence and keep the account.
Open tracking is the least reliable input you own. Privacy proxies fire it for people who never saw the message, so the branch triggers on nothing.
A branch that adds three near-identical emails is not a branch. It is a longer sequence wearing a diagram, and it fails for the same reason.
Every branch needs its own last step. A branch without one becomes a queue nobody closes, which is how people end up in two tracks at once.
Exit rules, and the failure called ghosting
Design the exit before you design the middle. Trish Bertuzzi named the failure in 2016 and the whole category has since forgotten it: the rep just trails off, and nobody ever decides the sequence is over.
- ✓A reply of any kind, including a no
- ✓An opt-out, however informally worded
- ✓A hard bounce, or an auto-reply naming a leaver
- ✓A not now that carries an actual date
- !No last step, so the sequence just fades
- !A break-up email written as a guilt trip
- !Ending the sequence and the account together
- !Opt-outs processed by hand, eventually
Write the last step first. It forces you to decide what the sequence was for, and every step in between gets easier to justify. Exits used to be manners. Under the current sender rules they are infrastructure, because a complaint costs you more than a lost prospect ever did.
Re-entry: coming back without starting over
Ending a sequence and ending the pursuit are two different events, and most teams conflate them. The second run needs a reason, and the reason has to be theirs.
Re-entry needs a new reason, not a new subject line. Circling back six months later is the same message with dust on it, and the reader can tell. What earns a second run is something that changed on their side, which conveniently also gives you a first line that could not have been written before. So when you exit an account, attach the reason you would come back for, not a date in the calendar.
Budget for the problem now exists, in public.
New plans, and a reason the old no expired.
The relationship traveled to a new account.
Working out which changes are worth a second run, and spotting them without a research habit, is the whole of signal-based selling.
Is it the sequence, the list, or the offer?
When a sequence produces nothing, three things can be broken and they have different symptoms. Check them in this order, because most founders start at the sequence purely because it is the easiest thing to edit.
- ·Bounces above a trickle
- ·Replies saying I don't handle that
- ·Near-zero opens on a healthy domain
- →Fix targeting and data, not copy
- ·Delivery and opens are fine
- ·People read, then say a polite no
- ·Two different angles fail the same way
- →Nothing in the sequence substitutes
- ·Opens with no replies at all
- ·A reply at step one, then silence
- ·Interest that dies while you were late
- →Fix spacing, jobs, channels, exits
Two different angles failing in the same way is the tell that the offer is the problem, and no amount of sequence design has ever fixed one.
Sizing the sequence to your mailbox estate
This is the part the last generation of sequence advice was written before, and it changes the design. You size the sequence to the estate, not the other way round.
Steps times people equals sends
Six email steps to 500 people is 3,000 sends. Do that multiplication before you pick a step count. A sequence you can only run at quarter speed is not a slower sequence, it is a worse one, because every gap stretches and the timing logic dies with it.
Divide the daily cap across the estate
Google's Workspace sending-limits page, updated 14 August 2026, caps a single user at 2,000 messages a day, applied over a rolling 24-hour window rather than resetting on a clock. That ceiling, divided across your estate, is the real constraint.
Count the domain, not the mailbox
Google Workspace Admin Help counts the 5,000-a-day bulk threshold across the primary domain, subdomains included: its own example is 2,500 from a domain plus 2,500 from a subdomain making you a bulk sender. The same FAQ says the classification does not expire. Checked August 2026.
Authenticate before you cross the threshold
Microsoft's own guidance on error 550 5.7.515 says senders of 5,000 or more messages a day to Outlook.com, Hotmail and Live must have SPF and DKIM both passing, publish DMARC, and align at least one with the From domain, or the mail is rejected.
Unsubscribe, and the complaint ceiling
Since 1 February 2024 Google has required one-click unsubscribe from bulk senders and told everyone to keep the spam rate reported in Postmaster Tools below 0.3%, recommending below 0.10%. Your exit rules now sit inside a published threshold.
Working out how many domains and mailboxes that actually takes is a separate exercise, and the cold email infrastructure tools we rate are where most teams end up. It is also the part we handle ourselves on an account: domains, inboxes and warmup, kept away from the main domain.
What to measure, and which reply counts
Two step-level numbers are enough for a two-person team. Most sequence dashboards show you six, and the four extra ones are decoration.
| What you measure | What it tells you | What it hides |
|---|---|---|
| Sends and touches logged | That the machine ran this week. | Whether a single person wanted any of it. |
| Open rate | Very little, now. | Privacy proxies open your mail for people who never read it. |
| Reply rate per step | Which steps are earning their place. | That a reply can be a no, or an angry one. |
| Positive replies per step | Whether the step creates conversations. Keep this one. | Nothing much. It is the closest thing to the outcome. |
| Opt-outs per step | Which step is a tax. Keep this one too. | Nothing. A step that only produces opt-outs is not persistence. |
We optimize for pipeline and replies rather than raw send volume, which is the same discipline at company scale: the number that flatters you is rarely the number that moves.
Multithread before you multichannel
If you can only add one thing to your sequence this quarter, add a second real contact at the account rather than a third channel.
higher response rate when the outreach reached several people at one organization instead of one.
A second contact is also the cheaper move: no new tool, no new mailbox, no warmup period.
Backlinko with Pitchbox, 2019, re-checked August 2026. The same analysis put the combined lift at 160% when several contacts ran alongside follow-ups. It skews to link-building outreach, so read the shape, not the level.
The per-persona detail, including who to add first and what to say differently, is in the multithreading play.
Running sequences as a two-person team
A sequence is not a thing you build, it is a thing you run. Maintenance is what kills it at this size, and nobody plans for maintenance.
-
1
Write the architecture down once
Aaron Ross and Marylou Tyler made this argument in Predictable Revenue in 2011 and it still holds: the sequence exists so the outcome does not depend on who is running it that week, or how they feel on a Thursday.
-
2
Book the weekly slot before you need it
One fixed hour, in the calendar, with a name on it. When the slot is the first thing to slip, the sequence keeps sending into a list nobody is refreshing.
-
3
Three tasks, same order, every week
Clear the reply queue. Retire the steps that only produced opt-outs. Top up the entry list. Anything else that week is optional, and these three are not.
-
4
Automate the sending, never the deciding
A tool can send on a schedule and stop on a reply. It cannot decide whether an account still fits or whether an angle has run out. Keep those two jobs on your side of the line.
We do this work on every account ourselves, with no junior team and no handoff after kickoff, so the weekly hour is not a delegated task for us either. The teams that keep a sequence alive are the ones that protect that hour when the week goes wrong, not the ones with the better tool.
A worked example: one sequence, designed from the rules
An illustrative walkthrough of the method, not a specific client result. Every step below is derived from the architecture rules on this page, not from a real account. We report real numbers only when they are real.
-
1Days 1 to 6 · The cluster
Three touches, two channels
Day 1 opens the reason. Day 3 adds what day 1 did not say. Day 6 moves to the other channel so a filter is not the last word. Tight on purpose: this week establishes that a person is here.
-
2Days 7 to 12 · The break
Nothing at all
Six days of designed silence. Nothing is sent, nothing is chased. The break is what stops the second half reading as pressure, and it is the step teams delete first when they get nervous.
-
3Days 13 to 20 · The stretch
Two touches, a smaller ask
Day 13 changes the ask from a meeting to a question about who owns this. Day 20 adds the second contact at the account. Gaps widen, so the tail never feels like the opening.
-
4Day 34 · The close
The step written first
One short message that says you are stopping and leaves the door on the latch. Then the account goes back to the pool with a re-entry reason attached, not a date.
Six touches over five weeks, because the ask is a first meeting with a cold buyer at an account that fits. Warm it up and the same rules produce four touches over two weeks. Aim it at a C-level buyer and they produce twenty over three months.
Where the common advice is wrong
Almost everything ranking for this topic is published by a company selling a sequencer or selling the service of running one. Five claims that do not survive being traced.
"Run 8 to 12 touches, keep going because most sales close after the fifth follow-up, and add a channel for a big multichannel lift."
- ✕8 to 12 touches is the number
- ✕80% of sales need five follow-ups
- ✕Multichannel lifts replies 2 to 4x
- ✕Replies flat, so rewrite the emails
- ✕Longer sequences build more pipeline
"Decide the ask and the temperature, size it to your send capacity, give every step a job, and design the exit first."
- ✓That average came from sellers surveyed about themselves
- ✓Nobody has ever located the organization it is credited to
- ✓No page quoting that range names a sample or a baseline
- ✓Check the list, then the offer, then the sequence
- ✓Most replies land on the first email, not the fifth
What to realistically expect
A well-built sequence buys you two things: the touch actually happens, and you are still welcome when it does. It buys nothing else.
Follow-ups roughly double your reply harvest and they do not rescue a first email nobody wanted. Backlinko with Pitchbox, 2019, found one follow-up lifting responses by 65.8%, and Instantly's own 2026 platform data puts 58% of replies on step one. Read together, those two say the same thing.
Architecture decides whether the touch happens. It never decides whether the offer is worth a meeting. A perfectly spaced sequence carrying a weak offer just delivers the rejection more efficiently, from someone who was briefly paying attention.
Qualified meetings, and a read on which segment and which message actually worked. When we run this for a team, that verdict takes about sixty days from kickoff. Design the sequence so it can produce that answer, which means one segment and one offer per sequence, and a log of what you sent to whom.
What to take away
- 1 Architecture decides whether the touch happens. Content decides whether it lands.
- 2 Every step needs a job. A step with no job is why people mute you.
- 3 Length follows the ask and the temperature, then your send capacity. Never a best-practice number.
- 4 Design the last step first, then the rules for leaving early.
- 5 When nothing works, check the list, then the offer, then the sequence. In that order.
Questions founders ask
How many touches should an outbound sequence have?
How far apart should the touches be?
What is the difference between a sequence and a cadence?
How do I know whether it is my sequence, my list, or my offer?
When should I stop emailing someone?
Is it true that 80% of sales need five follow-ups?
Do I need a sequencing tool to start?
Co-founder of Real Good GTM. He has been a first business hire and Chief of Staff at seed-stage B2B startups, building and rebuilding outbound sequences from an operator's chair rather than an agency script. He writes the architecture side of this site: how a motion is put together, and why it stops working.
Connect on LinkedInFrom the architecture to the actual sending
You have the shape. These three cover the channel decision, the tools that run it, and what happens the moment somebody replies.
LinkedIn vs cold email
Which one to start with, why the two are almost never compared honestly, and what each costs you to run.
Read the comparisonBest cold email tools
The sequencers compared on what they actually do to your architecture, including who is better off without one.
Compare the toolsReply automation
The queue that decides whether the sequence was worth running: replies classified, answered, and pushed to a call.
See the serviceWant this built and run for you?
Book a fit check. We'll look at your ICP, the offer you are taking to it, and whether outbound is the right motion for your stage at all. If it isn't, we'll tell you that too.
Book a Fit CheckNo hard sell. No fake numbers. Real good work speaks for itself.